Article: How the Calgary Flames are Laughing at Advanced Stats

[DJones]

I always find the minny, toronto and avalanche comparison to be weird. Sure Toronto has fallen spectacularly but that is mostly a hot goalie can trump everything else.

I'd be happy to be on Dallas's and Minnesota's path though. Dallas is having a horrid year due to bad defense, injuries, and even worse goaltending. If anything this is the freak season, not the ones where they did well.

Minnesota is Minnesota. Horrid goaltending is the only reason they aren't/weren't an elite team.

[DJones]

And Avalanche I think will be a very solid team next year. Just because Toronto did what Toronto does, it does not make that the rule.

And again as someone pointed out a while ago. The 2012 Ducks would also fall into that category. Not bad company to keep.

[CorsiHockeyLeague]

Originally Posted by Tinordi

McIndoe is a big advanced stats guy.
He was on a panel saying that the stats guys have basically won and that anyone in management who cares about their job in the NHL is now a stats guy.

He also wrote this, which is directly linked to from the article in the OP:

http://grantland.com/the-triangle/th...ics-awakening/

[_Q_]

Thing is, advanced stats don't really work for any team this season, which is probably why the Corsi bangers have shut up for the last month or so.

In terms of 5v5 corsi close:
Winnipeg, Boston and LA are in the top third
Dallas and Carolina are in the top half
Anaheim and NYR are in the bottom half (both of which are challenging for the president's trophy)
Vancouver is in the bottom half
Montreal (another potential president's trophy champ), Calgary and Ottawa in the bottom third.

[CorsiHockeyLeague]

11 of the top 15 teams in ES corsi are in the playoffs currently... 2 of the bottom ten are in: Calgary, and Montreal (on the back of maybe the best single season goaltending performance ever).

It strongly aligns, as usual, to who's in and who's out.

[GranteedEV]

Originally Posted by CorsiHockeyLeague

It strongly aligns, as usual, to who's in and who's out.

Aligns, but doesn't indicate. Correlates to, but doesn't predict. And it never accounts for roster turnover despite analytics experts always pointing to cross-season shifts.

The 2013 Wild are pointed to as a "fluke" team. Well are they still a fluke team?

The Flames becoming a better corsi team would follow if they likely become a stronger overall team. That's not gonna be denied by anyone.

The Flames staying pat as a corsi team does not predict them out of the playoffs next year. Specifically, it can't predict their shooting percentage to "regress".

Analytics are excellent until they're treated as predictors. That's why we can't stand people using Analytics as predictors of success.

[CorsiHockeyLeague]

Originally Posted by GranteedEV

Analytics are excellent until they're treated as predictors.

And... wrong.

http://dcr2-bc-sg810/?cfru=aHR0cDovL...9ja2VkLXNob3Rz

I've probably posted that link at least ten times now and the same arguments keep coming.

[neo45]

Originally Posted by CorsiHockeyLeague

And... wrong.

http://dcr2-bc-sg810/?cfru=aHR0cDovL...9ja2VkLXNob3Rz

I've probably posted that link at least ten times now and the same arguments keep coming.

Link isn't working for me

[heep223]

Originally Posted by GranteedEV

Aligns, but doesn't indicate. Correlates to, but doesn't predict. And it never accounts for roster turnover despite analytics experts always pointing to cross-season shifts.

The 2013 Wild are pointed to as a "fluke" team. Well are they still a fluke team?

The Flames becoming a better corsi team would follow if they likely become a stronger overall team. That's not gonna be denied by anyone.

The Flames staying pat as a corsi team does not predict them out of the playoffs next year. Specifically, it can't predict their shooting percentage to "regress".

Analytics are excellent until they're treated as predictors. That's why we can't stand people using Analytics as predictors of success.

Originally Posted by CorsiHockeyLeague

And... wrong.

http://dcr2-bc-sg810/?cfru=aHR0cDovL...9ja2VkLXNob3Rz

I've probably posted that link at least ten times now and the same arguments keep coming.

Originally Posted by heep223

Please, please don't turn this thread into another advanced stats debate.

Guys....please honour my OP

This article really captured the essence of the team for me. I'm so emotionally invested at this point that missing the playoffs will be a huge disappointment in the short term. In the longer term no matter what happens, this experience is going to be so beneficial in developing our young players.

Also I really like that Treliving has stated over and over that they're going to stay the course during the rebuild, and then backed it up at the trade deadline by accumulating more picks. I agree with the article that says a risk that the team faces is trying to "accelerate" the rebuild.

Going to the game tonight...GFG

[getbak]

Originally Posted by EldrickOnIce

If you think those three scoff at analytics, you haven't been paying attention.

Burke doesn't scoff at analytics. He does scoff at calling things like Corsi and Fenwick "analytics" though.

[DJones]

So possession stats seem to tell about 70% of the story and for predictive ability they assume past games have anything to do the next year.

I've followed economics long enough to realize they are just pulling #### out of their asses. They can explain the past very well but it's borderline useless in trying to predict the future.

[Vansmack]

Originally Posted by CorsiHockeyLeague

And... wrong.

http://dcr2-bc-sg810/?cfru=aHR0cDovL...9ja2VkLXNob3Rz

I've probably posted that link at least ten times now and the same arguments keep coming.

this link seems dead...

[_Q_]

Originally Posted by CorsiHockeyLeague

11 of the top 15 teams in ES corsi are in the playoffs currently... 2 of the bottom ten are in: Calgary, and Montreal (on the back of maybe the best single season goaltending performance ever).

It strongly aligns, as usual, to who's in and who's out.

No it really doesn't. 69% is definitely not a good success rate for a system to be considered predictive of anything.

As of right now the Kings are on top of the Corsi table. And while yes, they're good, they're not at the top of the league (that's why they're battling for a wild card spot).

Winnipeg at #8, Carolina at #9, Anaheim #17, NYR #19, Vancouver #20, Montreal #24 (one spot below Edmonton) and Calgary #28 (one spot below Toronto) all speak volumes regarding how terribly flawed corsi is at predicting anything useful.

Edit: 16 teams make the playoffs not 15. So corsi, this season has predicted 11/16 spots which is just under 69%

[PsYcNeT]

Originally Posted by _Q_

No it really doesn't. 73% is definitely not a good success rate for a system to be considered predictive of anything..

I don't have a horse in this race, but compared to many prediction systems, 73% is pretty damn good. Most barely beat a coin flip.

[_Q_]

Originally Posted by PsYcNeT

I don't have a horse in this race, but compared to many prediction systems, 73% is pretty damn good. Most barely beat a coin flip.

Sure, but in a league where more than half the teams qualify for the playoffs, flipping a coin will probably give you a better than 50% success rate if your goal is simply to figure out who does and who doesn't qualify for the playoffs.

My example above listed 8 fairly blatant examples of why the system is flawed. But there are probably close to 6 or 7 more teams where corsi doesn't really represent where the team is (Colorado at 29th, Arizona at 23rd, Edmonton at 22nd, St Louis at 11th, etc.) So really, how useful is the system when probably half the teams corsi standings don't align with the actual standings this far into the season?

[Tinordi]

How dare those people try to glean insight into performance by using statistics, the nerve!

[Cleveland Steam Whistle]

Question, is that 73% success rate the rate at which Corsi can predict the outcome of a game or season based on whatever feeds into it? I honestly don't know.

The reason I ask, is if that's the case, using Corsi success % or ratio's to figure out what's going to happen in the last 10 games is useless at this point. Because that same Corsi ratio would have suggested that the Flames would not be in the picture at this point. Meaning essentially, the Flames have overcome their Corsi over the course of 73 games, or said another way, Corsi has proven over the course of 73 games not to be a good predictor or outcome when it comes to the Flames, so there is no reason to believe that it would start to be applicable in the last 10.

Again, not sure I'm understanding the predictability ratio right, but that would be my thoughts.

I'm assuming the 73% is the confidence ratio at which Corsi can be deemed accurate or able to predict? If so, then that is the answer to this whole Corsi controversy crap. 73% is a fantastic directional predictor of outcome ratio. Meaning, if you were a gambling man, and wanted to predict the final standings of an NHL season at the start of said season, you'd likely get 73% of your prediction right. Can't really beat that predictability, in that regards.

But, from the perspective of statistical reliability, an essentially 30% error rate is terrible. It means that there are enough variants, issues or exceptions for the system that 30% of the time it doesn't apply. That's a huge number that simply can't be explained away as a "statistical anomaly". Which means, the Flames bad Corsi ratings aren't as likely to "catch up with them" as many would have you believe. Even a +/- 5% error margin wouldn't give you much hope other than to say that the Flames were defying odds this year with their record, and statistically it's bound to catch up with them eventually and they are going to pull a Colorado or Toronto. But a +/-30% error margin means something different. While it's still like the "best analysis tool" we have to try and predict with, it also means that there are enough fundamental flaws, or impactful details to outcome of games that the model doesn't capture that could be explaining the variance from the model. It doesn't have to just be a statistical anomaly that WILL even out over time, it could very realistically be that the team (in this case the Flames) are winning in ways that are sustainable but not captured by the model.

Anyway, Flames fans will obviously hope that the Flames success is due to them doing things not capture in the model (i.e. the 30% of error), while Flames detractors will hope the Flames actually should be falling into the 73%, but are part of a normal margin of error that that will eventually even out.

Wow, long ramble that I don't even know if I understand after reading it, but too much time spend now, I'm posting.

[_Q_]

Originally Posted by Tinordi

How dare those people try to glean insight into performance by using statistics, the nerve!

It has nothing to do with this.

[EldrickOnIce]

A team will go through hot stretches where they score lots and dry spells when they don't.
A goaltender will get and carry a team, and then cool off.
For the most part the game is cyclical like that.

'Corsi' tends to be more stable; in large part immune to random streaks or events.
It only makes logical sense that if you out shooting and outchance the opposition, you will tend to win more. It's a fools game to argue against that.

That's how I see the pro analytics argument as completely valid.

[EldrickOnIce]

Sort of double post..